Abstract
Purpose: :
To identify and assess the potential effects of selection bias in studies comparing strabismus surgery to botulinum toxin A (BTXA) injections for the primary treatment of esotropia (ET) in children.
Methods: :
MEDLINE and PubMed searches and review of references lists of related articles were performed to identify all English-language studies comparing surgery to BTXA as primary treatment for infantile or acquired ET in children. Study design and results were analyzed for sources of selection bias. Potential direction and magnitude of effect upon point estimates of effectiveness for each intervention were assessed.
Results: :
No randomized trials were found, and multiple non-comparative case series were subject to various selection biases. Only one comparative study was identified (J AAPOS 2010;14:111-116), a non-randomized cohort study of BTXA (1-3 bilateral injections) vs strabismus surgery for infantile ET in 442 children, reporting motor success rates of 45% and 69% (p<0.001), respectively. Bias and effects: (1) Allocation bias - surgery was more commonly recommended to older children, resulting in age differences between study groups. The effect was likely minimal, as age at treatment and duration of misalignment did not predict outcome. (2) Differential loss-to-follow-up bias - parents could opt out of BTXA after 1 or 2 injections but could not do so in the surgery group, potentially biasing the BTXA success rate downward. The effect on outcome was likely significant. 156 infants had surgery or no further treatment after only 1 or 2 injections and were treated as treatment failures. If these children had continued to receive BTXA injections, then based upon the reported marginal success rates for a 2nd and 3rd injection (35%), the overall BTXA success would be 62%, no longer different from surgery (p=0.79, Chi-square).
Conclusions: :
Differential loss to follow up biases the successful motor outcome rate for BTXA downward, likely by a significant degree. Adjusting for this bias resulted in equivalent success rates for surgery and BTXA. Future studies, even if randomized, may be subject to the same bias if withdrawal can occur prior to additional BTXA injections.
Keywords: strabismus: treatment • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology